Custom LLM and poor connection times.

Hi,

I created a chatbot using n8n as a backend, as I need access to my databases and tools.

Recently I discovered your agents/avatar system and I liked it. I want to use your avatar as a front-end, and my n8n chatbot as the back-end. I managed to get a good result using your Agent.speak function, as my chatbot manage my LLM interactions and the response is send to the speak function. I used your github code to implement the agent.

I have faced several problems:

1 When I call the agent.connect method I get this error message every time: "WebRTC: Using five or more STUN/TURN servers slows down discovery", and the process to get the stable connection may last 30 seconds! Thats too much for a client to wait. Is there anything I can do to improve the connection time?

2 I tried to use your Agent.chat instead, to try to improve the response time and to get the speech-to-text interface implemented. I have read that I can connect your agent to an external webhook, in order to use my own n8n llm implementation, but I haven't found anywhere the documentation to do it, or even an example. Could you tell if this is posible, and if so where to find the documentation?

Thanks in advance!